Whisper-Live

[Dimlight]

Is it possible to write a code for live automatic transcription like the following link: https://github.com/oliverguhr/wav2vec2-live?

It would be a great contribution.

[qunash]

I made a naive implementation which feeds audio from mic every 1 sec. It works okay.
https://huggingface.co/spaces/anzorq/openai_whisper_stt

[qunash]

I should add that this demo runs on a CPU, so it's super slow.
It's preferred to copy the code and run it on a GPU in a colab or on your local machine.

[olivMertens]

So next text incoming :) thanks

[olivMertens]

Try this morning,with english spoken and french, did'nt work...

[turnkit]

saw this today: https://github.com/shirayu/whispering

[zoq]

We worked on a live transcription, including a browser extension - https://github.com/collabora/WhisperLive

Also https://col.la/transcription has a demo for transcription, summary, and translation that is based on Whisper.